Add primitive docs for custom derivatives in Clad sphinx docs #1444
Conversation
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
vgvassilev
left a comment
vgvassilev
left a comment
There was a problem hiding this comment.
This looks quite promising.
@PetroZarytskyi, can you take a look and possibly review this PR, too.
@davidlange6, can you check English?
| Note that the constructor pullback does not need anything such as | ||
| :code:`clad::ConstructorPushforwardTag<::Coordinates>`. It is because | ||
| the constructor pullback takes :code:`d_coordinates` as an argument, which can be | ||
| used to identify the class for which the constructor pullback is defined. |
There was a problem hiding this comment.
Can we add a section on clad::zero?
There was a problem hiding this comment.
I think you meant clad::zero_init?
There was a problem hiding this comment.
I have mentioned why the constructor pullback custom derivatives does not have the initialization problem.
Currently, the docs does not cover object construction through custom derivatives. I would like to cover this topic as a follow-up PR.
|
|
||
| Constructor pullback custom derivatives are more similar to the ordinary pullback | ||
| functions. Constructor pullback functions do not have the same problem as of constructor | ||
| pushforward functions of initializing the primal object and the derivative object. After all, |
There was a problem hiding this comment.
Sadly, constructor_pullback doesn't have the same problem because the initialization must be done in the forward pass, and so this responsibility lies on constructor_reverse_forw. By the way, we need a section about that. We can often avoid using that by
A a(args);
-->
A a(args);
A da(a);
clad::zero_init(a);
But really, we can only avoid constructor_reverse_forw either for iterable containers, simple numerical structures (no pointer fields) like std::pair, or when the object is initialized with the default constructor. For complex types like std::vector, we rely on constructor_reverse_forw.
There was a problem hiding this comment.
I have mentioned why the pullback constructors does not have the same problem of initialization as of pushforward constructors.
Currently, the docs does not cover object construction through custom derivatives. I would like to cover this topic as a follow-up PR.
|
Thank you for the detailed reviews @vgvassilev and @PetroZarytskyi. I will implement the reviews soon. |
|
@arima-04, ping. |
arima-04
force-pushed
the
CustomDerivativeDocs
branch
2 times, most recently
from
ec8cbaf to
765fb95
Compare
arima-04
force-pushed
the
CustomDerivativeDocs
branch
from
765fb95 to
7e26619
Compare
Signed-off-by: Maki Arima <[email protected]>
arima-04
force-pushed
the
CustomDerivativeDocs
branch
from
7e26619 to
3b0133d
Compare
|
@vgvassilev @PetroZarytskyi Apologies for the delay. I have applied all the reviews suggestions. |
|
for english checking - could you first run the text through google docs and resolve the grammar suggestions that it has (it is likely at least as good as me) |
|
@arima-04, ping. |
I will get back to this PR tomorrow. I couldn't get a chance to run the documentation through google docs. |
4 participants
No description provided.